MRT Impact Analysis
Analysis Date: 2026-02-04
Data Period: 2021-2026
Primary Focus: HDB, with cross-segment comparison
Key Takeaways
The clearest finding
For HDB, the “MRT premium” is much smaller than popular market language suggests. Much of the apparent accessibility premium is better explained by CBD proximity and broader neighborhood context.
What this means in practice
- HDB buyers should not pay large premiums purely for shorter MRT distance.
- Condo buyers and investors should still take MRT seriously; condo pricing is far more transit-sensitive.
- Town context matters. MRT access does not price the same way across central and suburban locations.
Core Findings
1. MRT has a modest average effect on HDB prices
| Metric | Value | Interpretation |
|---|---|---|
| HDB MRT premium | $1.28 per 100m closer | Small average pricing effect |
| Mean HDB price PSF | $552 | Context for the premium |
| OLS R² | 0.52 | Moderate explanatory power |
On its own, that average is too small to justify many listing premiums seen in practice.
2. The average hides large town-level differences
| Town | MRT Premium per 100m | Mean Price PSF | Transactions |
|---|---|---|---|
| Central Area | +$59.19 | $903 | 599 |
| Serangoon | +$12.91 | $566 | 1,853 |
| Bishan | +$5.88 | $644 | 1,951 |
| Marine Parade | -$38.54 | $629 | 515 |
| Geylang | -$20.54 | $584 | 2,054 |
| Sengkang | -$16.88 | $558 | 7,585 |
Impact
- In central areas, MRT access can still command a meaningful premium.
- In some suburban and already well-served areas, shorter station distance may coincide with crowding, noise, or other offsetting trade-offs.
3. MRT matters far more for condos than HDB
| Property Type | MRT sensitivity |
|---|---|
| HDB | low |
| Condo | roughly 15x higher than HDB |
Impact
- Investors should keep separate heuristics for HDB and condo.
- A transport argument that makes sense for condo often does not translate well to HDB.
4. Other amenities can outrank MRT in HDB pricing
| Feature | Importance |
|---|---|
| Hawker within 1km | 27.4% |
| Remaining lease months | 14.1% |
| Park within 1km | 7.2% |
| MRT within 1km | 5.5% |
Impact
- For HDB, daily convenience and lease still matter more than station access in the pricing model.
- Buyers who stretch only for MRT may be paying for the wrong attribute mix.
Decision Guide
For HDB buyers
- Compare MRT distance against lease, town, and amenity trade-offs.
- Use town-specific evidence rather than a generic “near MRT is always better” rule.
For investors
- Prioritize MRT proximity for condos and future tenant demand.
- Be skeptical of HDB listings that justify large premiums mainly on MRT access.
For upgraders
- When moving from HDB to condo, recalibrate your accessibility assumptions. The same 300-500m difference carries different market value across segments.
Technical Appendix
Data Used
- Primary input:
data/parquets/L3/housing_unified.parquet - Sample: 97,133 HDB transactions, filtered to 2021 onward, with valid coordinates
- Spatial grid: H3 hexagonal grid at resolution 8 (~0.5 km² cells), minimum 10 records per cell
- CBD reference point: (1.2839, 103.8513), haversine distance calculation
Methodology
- OLS LinearRegression controlling for
dist_to_nearest_mrt,floor_area_sqm,remaining_lease_months - XGBoost (100 trees, max_depth=6, learning_rate=0.1) for non-linear patterns, with SHAP for explainability
- Hierarchical regression via
scripts/analytics/analysis/spatial/analyze_cbd_mrt_decomposition.py: CBD-only → CBD+MRT → Full model, to isolate incremental MRT contribution - VIF analysis to check multicollinearity between MRT and CBD distance
- Town-level OLS in
scripts/analytics/analysis/mrt/analyze_mrt_heterogeneous.py, minimum 500 transactions per town - Cross-property comparison in
scripts/analytics/analysis/mrt/analyze_mrt_by_property_type.py, interaction terms for property type × MRT distance
Technical Findings
- CBD-only model R² = 0.2263, CBD+MRT R² = 0.2341, ΔR² = +0.0078 (MRT adds less than 1 percentage point beyond CBD)
- Full model R² = 0.4977 (broader housing features dominate)
- HDB MRT coefficient: 552); Condo MRT sensitivity ~15× HDB
- Feature importance (XGBoost): hawker_within_1km = 27.4%, remaining_lease_months = 14.1%, park_within_1km = 7.2%, mrt_within_1km = 5.5%
- Town-level variation: Central Area +38.54/100m
- VIF: MRT and CBD distance show moderate collinearity (~0.5-0.7 correlation), confirming they capture overlapping but distinct location effects
Conclusion
The technical evidence supports the headline finding: CBD proximity is the dominant location factor for HDB, and MRT adds only a marginal incremental contribution. The ~15× condo-vs-HDB sensitivity gap is robust across OLS and XGBoost specifications. Town-level heterogeneity is large enough that a single national MRT premium is misleading. Key limitations: distances are straight-line to nearest station (not walking path quality), and some town coefficients likely absorb omitted neighborhood factors.
Scripts
scripts/analytics/analysis/mrt/analyze_mrt_impact.py— OLS + XGBoost + SHAP on H3 H8 gridscripts/analytics/analysis/mrt/analyze_mrt_heterogeneous.py— Town-level, flat-type, price-tier stratificationscripts/analytics/analysis/mrt/analyze_mrt_by_property_type.py— HDB vs Condo comparisonscripts/analytics/analysis/spatial/analyze_cbd_mrt_decomposition.py— Hierarchical regression, VIF, PCA